Start building. Scale when you're ready.
Every plan includes zero data retention, sovereignty in your jurisdiction, and the OpenAI and Anthropic interfaces you already know. Bring your own GPUs with a free account, or let us manage everything.
Free
For personal use, with private inference on your own hardware.
- 1 self-hosted Pendra Worker
- Rent on-demand GPU workers by the hour, pay as you go
- OpenAI- and Anthropic-compatible API
- Zero data retention
- Sovereign jurisdiction
- Python and Node.js SDKs
- Community documentation
Pro
For commercial use, for teams shipping AI products in production.
Everything in Free, plus
- £99 of credit to spend on on-demand GPUs
- Up to 5 self-hosted Pendra Workers
- Worker monitoring, power history and webhook alerts
- Standard DPA included
- Priority support by email
- Request logging, if you want it
- Private inference, encrypted end to end
Enterprise
Managed GPUs, advanced security and dedicated support.
Everything in Pro, plus
- Pendra-managed GPU infrastructure
- Unlimited self-hosted Workers
- Automatic data masking and redaction
- Granular audit logging
- Custom energy reporting
- Custom DPA and DPIA support
- Dedicated account manager
- SLA with uptime guarantee
- SSO and role-based access control
- Onboarding and integration support
Compare plans
| Feature | Free | Pro | Enterprise |
|---|---|---|---|
| Usage | |||
| Commercial use Use Pendra to power products, services or workloads for your customers or your business. | – | ||
| Infrastructure | |||
| Self-hosted Workers Run inference on your own GPU hardware, managed through Pendra. | 1 | Up to 5 | Unlimited |
| OpenAI- and Anthropic-compatible API Works with the OpenAI and Anthropic interfaces you already know. | |||
| SDKs Client libraries for Python and Node.js, with end-to-end encryption built in. | |||
| Pendra-managed GPUs Dedicated GPU clusters in UK data centres, fully managed by Pendra. | – | – | |
| Operations | |||
| Enhanced usage analytics Request volume, latency percentiles, model usage breakdowns and per-key tracking. | – | – | |
| Worker monitoring and webhook alerts A health dashboard, with alerts when a worker goes offline or errors spike. | – | ||
| Power usage How much power each worker's machine draws. Live on every plan; Pro adds history up to 30 days with energy used, average and peak. | Live | Live + history | Custom reporting |
| Optional request logging Opt-in logging of requests and responses. Off by default, and you choose what to capture. | – | ||
| Priority support Faster response times from the Pendra engineering team. | – | Dedicated | |
| Security and compliance | |||
| Zero data retention Prompts and completions are processed in memory and never written to disk. Architecture, not policy. | |||
| Sovereign jurisdiction Run on our UK compute or your own GPUs in your country. No foreign government, including under the US CLOUD Act, can compel access. | |||
| DPA A Data Processing Agreement for UK GDPR compliance, standard or custom-negotiated. | – | Standard | Custom |
| Private inference Prompts and completions are sealed on your side and only decrypted inside your worker. Pendra relays data it cannot read. | – | ||
| Automatic data masking Automatic redaction of personal and sensitive data before it reaches the model. | – | – | |
| Audit logging Granular logs of all API activity and configuration changes. | – | – | |
| DPIA support Input and documentation for your Data Protection Impact Assessments. | – | – | |
| SSO and role-based access Single sign-on and role-based access control for your organisation. | – | – | |
| SLA A contractual uptime guarantee with defined response and resolution times. | – | – | |
Frequently asked questions
Is my data stored or logged?
No. On every plan, prompts and completions are processed in memory and never stored. This is how Pendra is built, not a setting you have to remember.
Can I use the Free plan for my startup or product?
The Free plan is for personal use: evaluating Pendra, running experiments and hobby projects. Anything where Pendra powers a product, service or workload for your customers or your business needs Pro or above. If a side project earns revenue, that's commercial use.
What's a Pendra Worker?
A Worker is the Pendra agent that runs on your own GPU hardware and serves models through the same API as Pendra-managed infrastructure. It connects out to Pendra, so you don't open any inbound ports in your network. Requests reach your Worker through the Pendra API, where they're processed in memory and never stored.
What hardware do I need to run a Pendra Worker?
Workers run on Linux, macOS or Windows, with NVIDIA GPUs, Apple Silicon or CPU only. Install a model from the catalogue and the Worker serves it, with nothing else to set up. Memory needs depend on the model: a 7B model fits in 16 GB, and a quantised 70B model needs 48 GB or more. See the Workers docs for full requirements.
What models can I run?
Pendra serves open-weight models, including Llama, Mistral, Qwen, DeepSeek, Gemma, GLM, Phi and gpt-oss, alongside image generation (Flux), audio transcription (Whisper, Parakeet) and embeddings (Nomic, BGE). We don't offer proprietary models like GPT, Claude or Gemini. Open-weight models on sovereign infrastructure mean full transparency and no lock-in at the model layer.
How is Enterprise pricing structured?
Enterprise pricing is based on your GPU requirements, model selection and throughput. Managed GPU plans are flat-rate: dedicated compute at a fixed monthly price, with no per-token billing. Get in touch and we'll scope it with you.
Can I mix managed and self-hosted?
Yes, and it's common. Run most of your traffic on Pendra-managed GPUs while routing your most sensitive customers' requests to self-hosted Workers in their own environment, all through one API. Mixing the two is an Enterprise feature.
What compliance certifications does Pendra hold?
Pendra is UK GDPR compliant and Cyber Essentials certified, with Cyber Essentials Plus and ISO 27001 in progress. Pro customers are covered by our standard Data Processing Agreement; Enterprise customers get a custom DPA and Data Protection Impact Assessment (DPIA) support tailored to their deployment.
Where exactly does my data live?
Prompts and completions are processed in memory on the Worker and never written to disk. The only data we keep is operational metadata (timestamps, model names, token counts and latency), stored in the UK by Pendra AI Ltd. No content crosses a UK border at any point.
How does billing work?
Pro is billed monthly in GBP. There's no per-token charge: the plan price is what you pay. Cancel any time from the dashboard, and access continues until the end of the billing period. Enterprise is invoiced annually on custom terms.
Can I upgrade or downgrade at any time?
Yes. Move between Free and Pro at any time. Enterprise changes are handled with your account manager.
Not sure which plan fits?
Book a 15-minute call. We'll help you find the right setup for your workload and your compliance requirements.